Blog Post #1 – The Influence of The Clustering Coefficient on Word-Learning
Study: https://www.frontiersin.org/articles/10.3389/fpsyg.2014.01307/full#B6
Network science is often used in the study of relationships between vocabulary and the learning and retrieval of words. One measure that is especially implemented in these studies is the clustering coefficient (C), defined as the the probability that two randomly selected neighbors of a node are neighbors with each other. In the study of vocabulary, the clustering coefficient assesses the extent to which neighbors of a word in a network are connected with each other, and is used to predict the difficulty of word-learning and word-retrieval. Researchers posited that words with with a high C would be learned more easily and retrieved with more difficulty through their connections versus words with a low C.
Previous studies have shown that the stimulation of words with low C, meaning the target word’s neighbors only have a few interconnections, resulted in the “activation [of some of the target word’s neighbors] spreading back to the target word, [while] the remaining activation dispersed to the rest of the network,” and the stimulation of words with high C, meaning the target word’s neighbors are highly connected, caused the activation to remain mostly among the neighbors instead of spreading to the rest of the network (Goldstein and Vitevitch). This study, conducted by Rutherford Goldstein and Michael Vitevitch, used 32 participants to expand upon previous studies, and determine how high and low Cs affected the learning and retrieval of word and unrelated image pairs, hereafter referred to as word-nonobject pairs. The results of their study uncovered two very interesting findings. The first finding of Goldstein and Vitevitch’s study was that in learning word-nonobject pairs, the low C tendency to spread remaining activations throughout the network meant that the target pair did not get reinforced, and thus was more difficult to learn than high C pairs, whose recirculating activation between neighbors served to strengthen the target pair. The second finding was that in word-nonobject pair retrieval, low C pairs “stood out” more from their neighbors, due to the fact some of the activation was spread back to the target pair, and thus were easier for participants to rapidly and accurately retrieve compared to high C pairs, for which identifying the target pair was more difficult due to all the highly activated neighbors.
Studies such as Goldstein and Vitevitch’s apply concepts from our class to demonstrate their application to real world phenomena. For example in this study, researchers incorporated simple topics, such as the clustering coefficient, to investigate complex questions such as what affects word-learning and word-retrieval. In class, we were taught that the clustering coefficient is the mathematical representation of the triadic closure principle. The triadic principle has one of the most basic roles in the formation social networks, as it states that if two people have a friend in common, they are more likely to be friends. Studies such as Goldstein and Vitevitch’s not only implement these principles, but do so in a way that highlight the networks, such as the language network, that underlie all social networks to reveal just how extensive the impacts of the topics we cover in class truly are. This study caused me to reflect on processes that contribute to social networks, such as language, that are also affected by the clustering effect in their own way. Studies such as Goldstein and Vitevitch’s, highlight how networks are deeply embedded in everyday life, as the topics we have learned in class may be applied to many different layers of the global social network.